investigating hologram-based route planning
Post on 15-Oct-2021
3 Views
Preview:
TRANSCRIPT
1
To appear in Transactions of Geographical Information Science
INVESTIGATING HOLOGRAM-BASED ROUTE PLANNING
Sven Fuhrmann1)
, Oleg Komogortsev2)
and Dan Tamir2)
1) Department of Geography, Texas State University, 601 University Drive, San Marcos,
TX 78666, Email: fuhrmann@txstate.edu (corresponding author)
2) Department of Computer Science, Texas State University, 601 University Drive, San
Marcos, TX 78666, Email: ok11@txstate.edu, dt19@txstate.edu
Short title: Hologram-based route planning
Keywords: Geospatial Holograms, 3D Geovisualization, Eye-Movement Analysis, Route
Optimization, Spatial Thinking
2
INVESTIGATING HOLOGRAM-BASED ROUTE PLANNING
Abstract
It is often assumed that three-dimensional topographic maps provide more effective route
planning, navigation, orientation, and way-finding results then traditional two-dimensional
representations. The research reported here investigates whether three-dimensional spatial
mappings provide better support for route planning than two-dimensional representations. In a
set of experiments performed as part of this research, human subjects were randomly shown
either a two-dimensional or three-dimensional hologram of San Francisco and were asked to plan
a bicycling route between an origin and a destination point. In a second task, participants used
these holograms to identify the highest elevation point in the displayed area. The eye-movements
of the participants, throughout the process of looking at the geospatial holograms and executing
the tasks, were recorded. The eye-tracking metrics analysis indicates with high statistical level of
confidence that three-dimensional holographic maps enable more efficient route planning. In
addition, the research group is developing a new algorithm to analyze the differences between
participant-selected routes and a set of ―good routes.‖ The algorithm employs techniques used to
represent the boundary of objects and methods for assessing the difference between objects in
modern digital image recognition, image registration, and image alignment applications. The
overall goal is to create a theoretical framework for investigating and quantifying route planning
effectiveness.
3
1. Introduction
Spatial thinking is often a ubiquitous cognitive process that many humans do not explicitly
realize or take for granted. In everyday activities we describe travel routes to out-of-town
visitors, plan a trip to the supermarket, and search for a pleasant place to retire. Spatial thinking
has been described as a collection of integrated cognitive skills. The Committee on the Support
for Thinking Spatially (2006) refers to spatial thinking as a combination of three elements: 1)
concepts of space, 2) tools of representation, and 3) processes of reasoning. The concepts of
space and spatial reasoning allow humans to understand the environment around them, make
decisions, and learn how to solve spatial problems. Spatial representations allow humans to store,
represent, and retrieve spatial information. Spatial representations facilitate spatial learning and
thinking and are essential for understanding, remembering, reasoning, and communicating
spatial content and concepts.
Humans have numerous spatial representations at their disposal. Cognitive maps,
physical two and three-dimensional maps, as well as recently introduced digital two and three-
dimensional representations are just a few examples. Many three-dimensional, often interactive,
representations e.g., Google Earth and Microsoft Virtual Earth, have been developed to support
spatial thinking. Although there is no conclusive research on this topic; many researchers,
developers, and users believe that three-dimensional representations provide superior usefulness
and usability. The research reported here attempts to determine whether three-dimensional
spatial representations support better spatial thinking then two-dimensional mappings. We
specifically focus on route planning as one aspect of spatial thinking and investigate whether
three-dimensional topographic maps enable more effective route planning compared to two-
dimensional topographic maps. Our study uses geospatial holograms as three-dimensional
4
representation media. Geospatial holograms provide almost the same physical properties and
affordances as traditional maps. We created two geospatial holograms of downtown San
Francisco: one that features a two-dimensional topographic representation of the area and a
second hologram that displays the topography as realistic three-dimensional representation.
These geospatial holograms enable direct comparison of two different visualization modes and
allow focusing on route planning with two-dimensional vs. route planning with three-
dimensional representations. Our overall hypothesis is that eye-movement analysis can be
efficiently used to quantify route planning effectiveness and to compare route planning patterns.
In an experimental setting we applied eye-movement recordings to investigate the visual
cognitive processes of three-dimensional route planning.
In addition, our research involves the development of a new algorithm to analyze the
differences between participant-selected routes and a set of ―good routes‖ (i.e., routes selected by
experts). The algorithm employs techniques used to represent the boundary of objects and
methods for assessing the difference between objects in modern digital image recognition, image
registration, and image alignment applications. Our prolonged goal is to create a theoretical
framework for investigating the effectiveness of map-based route planning and additional map-
based spatial thinking skills. The findings of the current research significantly advance the stated
goal.
2. Geospatial Holograms
Since Dennis Gabor developed the first holograms in the late 1940s, holograms have been an
object of scientific research and public interest (Heckman 1986; Stea 1976; Benton and Bove
2008; Hariharan 2002). Holograms are mostly known for visualizing three-dimensional objects
5
for entertainment and eye-catching purposes in amusement parks and art exhibitions.
Nevertheless, despite the fact that the technology is over half a century old and well documented,
holography has not been incorporated with mainstream visualization technology. This is due to
multiple reasons including the high degree of difficulty and long length of time for production.
Additional reasons include high cost as well as inconvenient restrictions on features such as size,
color representation, and viewing angles. With recent development of advanced holographic
software and hardware, these limitations are starting to vanish (Benton and Bove 2008; Kasper
and Feller 2001).
Holograms utilize the physics of light diffraction to create optical illusions of solid three-
dimensional objects or scenes (Kasper and Feller 2001; Hariharan 2002). They use special
photographic materials that can encode and reconstruct three-dimensional visual information by
means of a diffraction or interference pattern that can recreate a light wave-front that would
actually be reflected from a three-dimensional scene or object. The encoding of a light wave-
front into interference patterns is a complex process beyond the scope of this paper. When
correctly illuminated, holographic interference patterns are decoded by the human physiological
system and a realistic three-dimensional scene appears before the human eye (Kasper and Feller
2001; Hariharan 2002; Heckman 1986).
One advantage of holograms is that users can view the representations without special
glasses, goggles, or tethered eyewear. Only a single point light source, e.g., unobstructed
sunlight, a standard LED flashlight, or a standard halogen light, is required to make the imagery
visible (figure 1). Initially, input data for holograms were limited to physical objects that are on the
same scale as the recording material. Nowadays, three-dimensional digital scanning technologies
or existing digital three-dimensional models and scenes can be used to generate the input data for
6
holograms. Furthermore, the holographic production process has mainly become digital and the
production time for holograms has decreased from several days to several hours.
--- FIGURE 1 ABOUT HERE ---
In comparison to traditional maps and two-dimensional electronic displays, geospatial
holograms have a low (1 mm) resolution on the surface, but offer extremely rich information
content because of their directional resolution. Two-dimensional media, such as photographs or
maps display the same information regardless of the viewing angle while geospatial holograms
can display different spatial information according to the user viewing angle. Modern geospatial
holograms can actually encode spatial information for over one million different viewing angles.
In theory entirely different data sets (e.g., topographic maps) can be presented at different angles.
Benton and Bove (2008) provide more information about latest holographic technologic
developments.
Geospatial holograms are a unique category of holograms with special geovisualization
requirements. The relatively low resolution of 1mm pixels (figure 2) requires (1) applying
sophisticated generalizations to geospatial information, (2) generating sufficient text labels and
general map elements, (3) avoiding strong height exaggerations that might cause blurred
imaging, and (4) optimizing the monochromatic color scheme for geoinformation representation.
―Geospatial holograms‖ can be defined as three-dimensional holographic spatial representations
of natural or human-built environments. These geospatial holograms are generalized, symbolic
depictions of an area and usually represent relationships and patters of geospatial objects and
7
phenomena. Currently, most geospatial holograms are of monochromatic nature, but
experimental true color geospatial holograms are available at a relatively high cost. The research
project reported in this publication applies green monochromatic geospatial holograms. True
color geospatial hologram design will be investigated in the future.
--- FIGURE 2 ABOUT HERE ---
3. Route planning
One important spatial thinking process is route planning, a daily mostly ubiquitously executed
task. Route planning includes generating a possible path between an origin and a destination
point. It is an initial part of the way-finding process that humans usually accomplish in familiar
and unfamiliar terrain; using mental, analog, and digital spatial representations. Many spatial and
non-spatial factors influence route determination. For example, bicyclist might chose a route
with lower slope, but longer distance while a hiker might select the shortest distance without
worrying about slope. Overall, route planning and route decision making involves subjective
spatial thinking including cognitive mapping, as well as the use of objective external spatial
representations.
As humans become spatially literate they develop spatial knowledge and cognitive
processing abilities which enable succeeding in way-finding and other spatial problem solving
processes. Downs and Stea (1973) identify the mental representation building process as
cognitive mapping and the product as a cognitive map (Tolman 1948). The cognitive map is not
a map-like, single scale, distortion-free representation (Tversky 1993; Kuipers 1983; Hirtle
8
1997). It is a personal mental representation of space that includes various media information
(language, images, etc.) such as perspectives, reference points, and errors. (Tversky 2000, 1993;
Kitchin and Freundschuh 2000).
Human spatial knowledge is assumed to consist of landmark knowledge, route
knowledge, and survey knowledge (Blades 1991; Siegel and White 1975). Landmarks are spatial
features that are used to locate oneself in the environment. They act as guides during way-finding
(Blades 1991; Sorrows and Hirtle 1999). Route knowledge (procedural knowledge) enables
following routes and describing sequences of landmarks (Siegel and White 1975; Cornell and
Heth 2000). It is assumed that a sequence of decision factors (distances, turning points, etc.) is
stored and supports getting from one place to another (Kuipers 1978; Thorndyke and Goldin
1980). Survey knowledge (also referred to as configurational knowledge) allows the
understanding of spatial relationships (i.e., topology) in an environment (Siegel and White 1975;
MacEachren 1991). Freundschuh (2000) contends that long exposure to an environment leads to
survey knowledge and Thorndyke and Hayes-Roth (1980); Péruch et al. (2000) state that route
and survey knowledge can be acquired directly from maps. Cognitive representations derived
from maps, however, have been reported to be orientation-specific, while cognitive maps derived
directly from environmental experience appear to be orientation-free (MacEachren 1992; Tlauka
and Wilson 1996; Shepard and Hurwitz 1984). Billinghurst and Weghorst (1995), Péruch et al.
(2000), Ruddle (2000) and Biocca (1997) report that cognitive mapping occurs while traveling
through real environments and through the use of spatial products such as maps, photographs,
verbal descriptions and virtual environments.
9
Cognitive mapping is the basis for way-finding, which allows reaching and
recognizing a destination in a familiar or unfamiliar environment (Arthur and Passini 1992;
Peponis et al. 1990). Downs and Stea (1977) describe the way-finding process in four steps:
orientation: knowing, or understanding, location, and the spatial relations between current
and target location.
route planning: choosing a route that leads to the destination.
route monitoring: monitoring the route taken to confirm that it is the correct route, going in
the right direction.
destination recognition: recognizing that the correct destination or a nearby point, has been
reached.
Gärling et al. (1986) integrate these steps into an information processing model for human
way-finding. They argue that, at the beginning of a way-finding situation, a destination is
decided upon and localized using information obtained from the cognitive map and other media,
e.g., a topographic map. Thereafter, the selection of a route to the destination takes place and the
travel plan is executed. Changes in the travel plan might be necessary during execution,
depending upon trip length, the complexity of the environment, and the familiarity of the subject
with the environment. Upon arrival, the way-finding process is terminated (figure 3). This
research does not address the full way-finding process. It focuses on the route planning process
and evaluates the effectiveness of the process.
--- FIGURE 3 ABOUT HERE ---
10
4. Eye-Movement Analysis
4.1 Studies in Cartography and Geovisualization
Observing eye movements started in the early 19th
century when scientists tried to scrutinize the
types of information that the eyes absorb while moving (Dodge 1900). Later, eye movement
recording developed into an important research method that is often employed in studies
concerning understanding of reading patterns (Rayner 1998), usability studies (Poole and Ball
2004; Tamir et al. 2008), psychology (Field et al. 2004), and visual inspection (Vora et al. 2002).
Generally, humans direct their vision center, the fovea, towards objects and scenes they
would like to observe. At a high level of abstraction, eye movements can be separated into
saccades, (fast movements of the eye), fixations (a static and focused position of the eye), and
pursuits (smooth following of a ―slow‖ moving object) (Duchowski 2007). Detailed research
connects eye movement patterns to various brain related activities including spatial memory and
cognitive processes (Leigh and Zee 2006). Researchers assume that long fixation duration
demonstrate high cognitive loads, great depth in cognitive processing, and sometimes, difficulty
in problem solving (Castner and Eastman 1985; Duchowski 2007).
In the early 1970‘s cartographers initiated investigations on the use of eye-movement
analysis for understanding the visual cognitive processes of map reading and the impact of map
complexity on that process (Steinke 1987; MacEachren 1995). The majority of the eye-
movement studies in cartography were conducted in the 1970s and 1980s (Antes et al. 1985;
Castner and Eastman 1984, 1985; Dobson 1977, 1979), yet only few studies were published in
recent years (Heidmann and Johann 1997; Johann and Heidmann 1996; Bollmann, Johann, and
11
Heidmann 1999). Most cartographic eye-movement studies are concerned with questions about
how maps might be read, and researchers are analyzing individual scan paths. These scan paths,
however, are individualistic, especially if the study participants do not have a specific task to
follow. Steinke (1987) describes these studies as ―let see what happens when we place a map in
front of somebody‖. Castner and Eastman (1984) conclude that map interpretation includes two
factors: 1) graphic factors from the physical and Gestalt properties of mapping products, and 2)
cognitive factors relating to the actual map interpretation activities. Instead of focusing on scan
paths, Castner and Eastman (1984) suggest investigating the number of fixations, fixation
durations, and inter-fixation distances. They conclude that researchers should ask how the eyes
move and not where the subject looks. Unfortunately, eye-movement research declined in the
mid 1980‘s a development that Steinke (1987) describes as: cartographers becoming doubtful
and skeptic towards this kind of methodological approach. Recently, with emerging less-
expensive eye-tracking hardware and software, new studies in eye movement recordings,
specifically in geovisualization, are conducted and published (Swienty et al. 2008; Fabrikant et
al. 2008).
4.2 Detection of Basic Eye Movement Types
Eye-movement observations can be processed to extract and analyze three basic eye-movement
types: fixations, saccades, and pursuits. Fixations are eye-movements that keep an eye gaze
stable with regards to a stationary target providing visual images with high visual acuity.
Saccades are very rapid eye movements where the eye rapidly shifts from one fixation point to
another fixation location. A pursuit stabilizes the retina to a moving object and follows the object
smoothly. It is important to define the exact detection algorithm for eye-movement analysis,
12
because different parameterizations of an algorithm might lead to different results (Komogortsev
and Khan 2009). In the context of geospatial holograms the overall image properties are static,
thus the pursuit movement observations can be ignored.
The detection algorithm applied in this research is based on the Velocity-Threshold
Identification (I-VT) model (Salvucci and Goldberg 2000). A velocity-based classification
approach has more synergy with oculomotor mechanics due to the fact that all eye rotations are
measured in degrees rather than in a dwell time classification approach (Salvucci and Goldberg
2000). An eye position sample is classified as (1) belonging to a saccade if the calculated eye
velocity exceeds 30°/s and (2) a fixation if the eye velocity is less than or equal to 30°/s. The
threshold of 30°/s is justified by Leigh and Zee (2006). A saccade is defined as a sequence of
continuous eye position samples with calculated velocity exceeding 30°/s. The saccade
amplitude is defined as the Euclidean distance between the coordinates of the first and last point
in that sequence. Saccades with amplitudes of less than 2° and saccades where the eye tracking
failed to detect an eye position (even for a single sample) are discarded from analysis. A fixation
is defined as a sequence of eye position samples with a velocity of less than 30°/s where the
angular distance between samples does not exceed 2°. One parameter of interest is the fixation
duration where in addition to actual fixation, sequences of eye tracking failures and blinks with
less than 75ms are considered a part of the fixation duration. In our experiment, the minimum
fixation duration is set to 100ms. A fixation location is defined as the average over all the valid
eye-position coordinates, excluding micro saccades (saccades with amplitude of less than 1°).
The eye fixation duration is defined as a time difference between the last and first samples in the
fixation sequence.
13
Based on the algorithm applied, the following eye-tracking metrics are calculated for data
analysis:
Average saccade amplitude:
A large saccade amplitude (measured in degrees) indicates the ―consumption‖ of
meaningful visual cues (Goldberg and Kotval 1999). Therefore, large saccade
amplitudes in geospatial representations might indicate effective representations.
Average fixation duration:
Generally, long average fixation duration indicates a high cognitive load. Long
average fixation durations can be interpreted in two different ways as either: 1)
the user has difficulties extracting information or 2) the user is more engaged with
interpreting a representation (Just and Carpenter 1976). Hence, distinguishing
between the two is case specific.
Number of saccades:
Generally, high number of saccades indicates that significant amount of
searching takes place and therefore less time is allocated to efficient task
completion (Goldberg and Kotval 1999).
Number of fixations:
In general, a high number of fixations is indicative of a non efficient visual task.
Many variables can influence the number of fixations, one of them could be poor
representation designs (Goldberg and Kotval 1999).
Saccade/fixation ratio:
This figure describes the ratio between search activity (represented by the number
of the saccades) and processing activity (represented by the number of the
14
fixations). A small saccade/fixation ratio indicates that the user is spending more
―cognitive resources‖ on the task (higher cognitive load) and less cognitive
resources on gathering important background information (Goldberg and Kotval
1999). This particular metric is especially useful for eye movement recordings
with low data validity rates.
Pupil diameter:
Eye tracking systems enable measuring biometric data such as pupil diameter.
There is evidence that suggests that pupil size increases when subjects are asked
to process information. Likewise memory engagement, attention concentration,
and comprehension of complex sentences might increase the pupil size (Andreassi
2006; Beatty 1982). Some researchers suggest that high pupil size might be
indicative of a high cognitive load (Poole and Ball 2004).
Total eye-path traversed:
The amount of effort expended by the Human Visual System (HVS) while
completing a task is an important metric for the specific geospatial representation.
Ideally, the effort expended by the HVS can be represented by the amount of
energy spent by the HVS during the task. The energy expanded is dependent on
the amount of eye movements, the total eye-path traversed, and the amount of
force exerted by each individual extraocular muscle force during each eye rotation
(Tamir et al. 2008). The extraction of the individual extraocular muscle force is an
extremely complex task (Komogortsev and Jayarathna 2008). Therefore, in this
manuscript we are using the total eye-path traversed as approximation for the
15
effort. This metric, measured in degrees, represents the total distance traversed by
the eyes between consecutive fixation points during a task.
Time on task
Time on task (TOT) is not an eye movement metric. Nevertheless, in this research
the TOT is obtained from the eye tracking device by measuring the total time that
the HVS is engaged with the task. TOT is another indicative of effort and
effectiveness.
5. Effort Based Measurements of Usability
We are currently performing additional research were we establish relations between user
interfaces, user effort, and usability. Usable applications require less user interaction effort in
general. Thus, the current model estimates user mental and physical effort based on
measurements of mouse, keyboard, and eye-movement activities. Geospatial holograms, and
other geovisualization technologies, can be considered as user interfaces to geospatial and non-
spatial information. Hence, we are in the process of transferring the effort-based usability model
to our spatial thinking assessment framework.
Usability is a term that is broadly defined (Grady 1992; McCall et al. 1977; Nielsen
1993). The ISO/IEC 9126 standard defines usability as ―the capability of a software product to
be understood, learned, used, and attractive to the user when used under specified conditions‖
(Software Engineering-Product Quality-Part 1: Quality Model 2001), with the following
characteristics: understandability, learnability, operability, attractiveness, and compliance.
Another international standard, the ISO 9124, uses a slightly different denomination for
usability; it lists efficiency, productivity, and satisfaction as the main factors of usability. This
16
standard calls for evaluating the effort invested by the user as a part of the evaluation procedure,
yet the actual definition of effort as well as ways to measure it, are not well defined. In practice
researchers measure time on task and evaluate user activity; observations that might provide
limited results for determining the mental effort and effectiveness of users working with
geospatial holograms.
The premise of the research reported in (Tamir et al. 2008) is that there is a relationship
between usability and effort. Let 𝐸 denote the total effort, mental and physical, required to
complete a task, as defined by the following equations:
𝐸 = 𝐸𝑚𝑒𝑛𝑡𝑎𝑙
𝐸𝑝ℎ𝑦𝑠𝑖𝑐𝑎𝑙
𝐸𝑚𝑒𝑛𝑡𝑎𝑙 = 𝐸𝑒𝑦𝑒𝑚𝑒𝑛𝑡𝑎𝑙
𝐸𝑜𝑡ℎ𝑒𝑟 _𝑚𝑒𝑛𝑡𝑎𝑙
𝐸𝑝ℎ𝑦𝑠𝑖𝑐𝑎𝑙 =
𝐸𝑚𝑎𝑛𝑢𝑎𝑙 _𝑝ℎ𝑦𝑠𝑖𝑐𝑎𝑙
𝐸𝑒𝑦𝑒 _𝑝ℎ𝑦𝑠𝑖𝑐𝑎𝑙
𝐸𝑜𝑡ℎ𝑒𝑟 _𝑝ℎ𝑦𝑠𝑖𝑐𝑎𝑙
Most of the terms used in the equations are self explanatory and denote types of efforts
required for task completion. On the other hand, 𝐸𝑜𝑡ℎ𝑒𝑟 _𝑚𝑒𝑛𝑡𝑎𝑙 and 𝐸𝑜𝑡ℎ𝑒𝑟 _𝑝ℎ𝑦𝑠𝑖𝑐𝑎𝑙 respectively
denote the amount of mental and physical effort that cannot be represented through eye-tracking
and other observation methods. They can be considered as an error term that accumulates the
errors inherent in the logging and tracking along with the fact that there are other forms of
mental and physical effort that is not be recorded.
Precise methods for measuring mental effort (𝐸𝑚𝑒𝑛𝑡𝑎𝑙 ) are still in a theoretical stage.
Researchers have made progress measuring mental or cognitive activities using Magnetic
17
Resonance Imaging Systems. Another approach, still in the theoretical stage, is to measure
cognitive activities using eye tracking analysis. Methods for estimating physical effort 𝐸𝑝ℎ𝑦𝑠𝑖𝑐𝑎𝑙
are more precise. In the case of interface usability, it is possible to log subject‘s activities such
as number of mouse clicks, number of keyboard clicks, and number of pixels traversed by the
mouse thereby describing the manual effort (𝐸𝑚𝑎𝑛𝑢𝑎𝑙 _𝑝ℎ𝑦𝑠𝑖𝑐𝑎𝑙 ). Tracking eye movements with an
eye tracking device provides a method for making a precise measurement of eye mental effort
(𝐸𝑒𝑦𝑒𝑀𝑒𝑛𝑡𝑎𝑙 ) and eye physical effort (𝐸𝑒𝑦𝑒 _𝑝ℎ𝑦𝑠𝑖𝑐𝑎𝑙 ).
Lessons learned from this approach will be applied in future geovisualization research on
effort-based measurements of usability and effectiveness, and provide a deeper insight into
spatial thinking processes.
6. The experiment
6.1 Subjects
Thirty-eight undergraduate and graduate students from the departments of Geography,
Psychology, and Computer Science volunteered for the experiment. Four students disqualified
due to eye-tracking device failures. The remaining subject pool consists of 23 male and 11
female participants. The participant age ranges between 19-54 years with an average of 29.5
years (SD 9.6). Eight participants have no previous map interpretation experience, fourteen
subjects describe their map interpretation skill level as moderate, and twelve participants
consider themselves as experts in map interpretation.
18
6.2. Geospatial Holograms of San Francisco
A section of downtown San Francisco, CA serves as the study area for this research. Two
monochromatic geospatial holograms with height of 590 millimeters and width of 833
millimeters represent a topographic map of that section (figure 4 and 5). Both geospatial
holograms are at a scale of 1:5500, using 1mm holographic pixels and identical monochromatic
color schemes. The difference between the two geospatial holograms, however, is that one
hologram displays the elevation information in three-dimensions to the observer, while the
second hologram is a ―flat‖ two-dimensional hologram that does not convey elevation
information directly. Like any other topographic map, the elevation information is encoded
through contour lines in both geospatial holograms.
--- FIGURES 4 AND 5 ABOUT HERE ---
6.3 Procedure
Half of the qualified participants (17 out of 34) work with the two-dimensional hologram. The
second half of the subjects utilizes the three-dimensional geospatial hologram. The hologram of
San Francisco, covered behind a black curtain, is placed on a wall above a desk. The estimated
distance between the subject‘s eyes and the hologram is 1140 millimeters. Five markers, placed
on the wall, are used to calibrate the eye-tracking device. A subject has to complete two tasks.
The first assignment, task (1), is planning a bicycling route between an origin and a destination
point using the hologram. No time limit is given for route planning. After planning the route, the
19
subject draws the planned route on a paper map. The second assignment, task (2), consists of
using the hologram to identify the highest elevation in the landscape. After identifying the
highest point, the participant marks that point on a paper map.
Eye movement recordings are performed using a Tobii x120 eye-tracker with the
following characteristics: sampling rate - 120Hz, accuracy 0.5°, spatial resolution 0.2°, and drift
0.3°. Tobii x120 eye-tracker is non invasive and allows for 30x22x30 cm freedom of head
movement. Nevertheless to improve the accuracy of recording a chin rest is employed to fixate
subjects head. Third party and own software is used to extract metrics from the recordings.
Each participant reports to the study individually, is briefed on the general scope of the
study, is introduced to the workings of the eye-tracking device, and examines a hologram of a
different urban area. Next, the participant is seated in a desk chair across from the covered
hologram with his or her chins resting on a chin-rest. Once the eye tracking device is calibrated,
the participant is ready to start the experiment. After lifting the curtain the subject‘s eye-
movements and time to complete the task are recorded. Eye-tracking is stopped when the
participant indicates that he or she finished the route planning task. The participant draws the
selected route on a paper map and moves to the second task of identifying the highest elevation
in the landscape. Again, the subject‘s eye-movements and time to complete the task are recorded
until the participant indicates that he or she has identified the highest point. At this time, the
participant marks the highest point on a paper map.
7. Results
20
The study includes two tasks: planning a bicycling route between two points (task1) and
identifying the highest point in an area (task 2). It compares eye-tracking metrics of subjects
executing the tasks while using a two-dimensional geospatial hologram with eye-tracking
metrics related to the execution of the same tasks via a three-dimensional geospatial hologram
(figure 6).
Micro saccades, i.e., saccades with amplitudes of less than 1°, are ignored. Furthermore,
corrupted saccades, that is, saccades in which the eye tracking equipment fails for a part of the
duration of a saccade are excluded. Data recordings with validity of less than 70% are removed
from the metrics analysis. The results and conclusions presented here need to be considered as
being preliminary, since additional studies are necessary to confirm these findings.
--- FIGURE 6 ABOUT HERE ---
The measurements of the average saccade amplitude (in degrees) with the three-
dimensional geospatial hologram yield a mean of 4.32 with standard deviation of 1.17 (M = 4.32,
SD = 1.17). The results of the average saccade amplitude with the two-dimensional geospatial
hologram are (M = 3.42, SD = 0.76). The statistical significance of these results is: F(1, 33) =
5.7, with a level of confidence: p = 0.02. On average the saccade amplitude for the three-
dimensional hologram is larger than for the two-dimensional hologram. These results indicate
that three-dimensional hologram users might be able to efficiently move their eyes between
distant points. The two-dimensional saccade amplitude results indicate that observers often move
their eyes between closer located points. It seems that – based on saccade amplitude observations
21
– the three-dimensional geospatial hologram would support more efficient route planning than
the two-dimensional counterpart.
The average fixation duration (in milliseconds) is (M = 485, SD = 173) with the three-
dimensional hologram and (M = 753, SD = 697) with the two-dimensional hologram. In this
case, F(1, 33) = 3.11, and p = 0.087. The longer fixation duration for the two-dimensional
representation indicates that this representation might require a higher cognitive effort, possibly
due to the less salient three-dimensional information embedded in two-dimensional spatial
representations.
Another set of results is the ratio of the number of saccades to the number of fixations
(the saccade/fixation ratio). The recorded ratio for the three-dimensional geospatial hologram (M
= 0.88, SD = 0.31) is significantly lower than the ratio for the two-dimensional geospatial
hologram (M = 1.21, SD = 0.2). The results significance is F(1, 33) = 10.9 and p = 0.002. These
results indicate that, in general, three-dimensional hologram users might experience a smaller
number of saccades and a larger number of fixations. It seems that users of three-dimensional
holograms do not move their eyes as often as two-dimensional hologram users. This finding
could be a good indicator for efficient route planning with three-dimensional geospatial
holograms.
The average pupil diameter is larger for the three-dimensional geospatial hologram (M =
4.3, SD = 0.4) than for the two-dimensional geospatial hologram (M = 3.9, SD = 0.78). The
difference between the results, however, is on the verge of statistical significance (F(1, 33) =
3.87, p = 0.057). This result might indicate that three-dimensional geospatial holograms provide
22
more information to accomplish a route planning tasks, thus might require a higher cognitive
processing load – a finding that needs to be further assessed.
The route planning task yields a higher time on task for the three-dimensional geospatial
hologram (M = 33, SD = 20.4) than for the two-dimensional geospatial hologram (M = 23.6, SD
= 15). The difference between these measurements is not statistically significant (F (1,14) = 0.85,
p = 0.317).
The traversed eye-path for the route planning task is longer for the three-dimensional
hologram (M = 300, SD = 173) than for the two-dimensional counterpart (M = 158, SD = 69).
The difference in results for the traversed eye-path is also not statistically significant (F
(1,14)=3.08, p=0.101). The latter two results might not necessarily indicate less effectiveness in
route planning. In the case of the holographic representation and the novelty associated with it
we can not rule out an ―awe‖ effect. The results may also denote that a three-dimensional
representation might contain more information to process and users might take the time to
explore different routing options to find ―their‖ best route – a finding that will further
investigated.
8. Discussion
Eye-movement analysis allows to investigate one aspect about effective route planning with two
and three-dimensional spatial representations. Steinke (1987) asserts that a combination of
methods is required for understanding the cognitive processes in map interpretation. The eye-
movement analysis sheds light on the effectiveness of three-dimensional spatial representations
and indicates that three-dimensional maps might help performing efficient route planning. The
23
eye-movement analysis, however, does not reveal information related to the quality of the
planned route. In other words, it does not indicate whether users of the three-dimensional
geospatial hologram plan better routes than users of the two-dimensional counterpart.
In order to investigate the initial findings of the eye-movement analysis we are developing a
new algorithm that utilizes the image processing concepts and shape features to analyze the
differences between participant-selected routes and a set of ―good routes‖. In image processing,
difference is measured using theoretical ―distance‖ measures such as the Mahalonobis distance
(Haralick and Shapiro 2002). Several image processing algorithms utilize shape features in the
process of object matching and use distance between object contours for object matching (Brown
1992). This procedure can be utilized for identifying the distance (i.e., the difference) between
routes. The following section explains the proposed ―length code matching algorithm‖.
8.1 The Length Code Matching Algorithm
Object shape features such as chain codes, Fourier descriptors, and object-signature are often
used for object matching and image registration (Freeman 1974; Haralick and Shapiro 2002). A
desirable characteristic of many of these features is their set of invariants which can provide
matching robustness. For instance, chain-codes are invariant to translation. A commonly used
set of shape features is referred to as the object signature (Haralick and Shapiro 2002).
The term object-signature is ―overloaded.‖ In the context of this paper it is the set of
distances measured from a fixed-point referred to as the ‗center‘ (generally this point is centroid
of the object) to pixels that reside on the object boundary (contour). In order to extract an object
signature the image has to be segmented, objects have to be identified, and the pixels that reside
24
on object contours have to be marked (Alajlan et al. 2006; Brown 1992). In our research, routes
are drawn by participants on paper maps and manually digitized.
We propose using a variant of the object signature referred to as the ―length code‖. The
main advantage of the length code is that it is less sensitive to ―noise‖. This is due to the fact that
a small change in shape, which may be due to noise, causes minimal changes to the distance
between the centroid and contour points. In addition, length coding is invariant to the entire set
of affine transformations. Finally, this representation converts a two-dimensional problem into a
one-dimensional problem without loss of information and provides a computationally efficient
framework.
Generally, in image processing algorithms for object matching, the contours of objects are
closed, i.e. the origin and the termination point are identical. The routes obtained in our eye-
movement analysis experiment are generally ―open‖ lines. They start in a point A and end in a
distinct point B. Lines pose only a minor difference to the image processing algorithm, since an
arbitrary reference point can be used.
The cyclic auto correlation function of the signature sequence is rotational and translational
invariant (Baggs 1997; Haralick and Shapiro 2002). The first autocorrelation coefficient of the
signature, 0R , is the sum of squares of the values of signature elements. Hence, it approximates
the energy of the signal and is proportional to the area of the object. Thus, in order to normalize
the sequence each autocorrelation coefficient iR is divided by the coefficient 0R . The resultant
sequence is scale invariant. It is referred to as the ―length code‖ of an object. The length code is
invariant with respect to translation, rotation, and scaling. Hence, it is invariant with respect to
affine transformations.
25
An important contribution of the proposed algorithm is the usage of dynamic space warping
(DSW), a dynamic programming (DP) algorithm, in the process of matching the length-code
features of tours. In several pattern recognition applications, L1-norm, or L2-norm are used as
the distance (or dissimilarity) measure. These measures work well under the assumption of
linearity and for well-formed contours. When the patterns are subject to non-linear
transformation and distortion due to different acquisition parameters, the L1 and L2 norms may
fail to produce accurate results. This is actually the case for the routes designed by the subjects in
the geospatial hologram tests. Each subject is drawing a route from a starting point to a
destination point, and the set of routes can be considered as a non-linear transformation of the
ideal route. A non-linear distance measure that can be considered in this case is the dynamic time
(or space) warping function. In this research, the DSW algorithm returns a measure of the
distance between two length-code sequences. The unique property of this distance measure is
that it allows for spatial fluctuations in the object (routes in our case) representation and
compensates for imperfections in digitization and shape representation.
Using experience obtained in object matching research (Baggs 1997; Baggs and Tamir
2008; Keogh et al. 2006), we are currently developing a DSW and length code based matching
algorithm for the evaluation of routing quality. The algorithm and the findings of the research
will be reported elsewhere.
9. Outlook
The international research agenda (MacEachren and Kraak 2001) of the Commission on
Geovisualization, International Cartographic Association encourages scientists to further
26
investigate the contexts within which geovisualization is successful. MacEachren and Kraak
(2001) pillory repeated statements in geovisualization publications that provide common-
knowledge claims about usefulness and usability of geovisualization without scientific proof.
The research presented here investigates the usefulness and usability of three-dimensional
representations over two-dimensional spatial representations in route planning and way-finding
situations. Route planning is a ubiquitous spatial thinking process. Evaluating this process
requires sound theoretical foundations. For this end, we have developed a theoretical framework
that is based on eye-movement analysis and measuring similarity between routes through a
length code matching algorithm to analyze route planning outcomes. Overall, the eye-movement
analysis provides important and promising results on route planning effectiveness with three-
dimensional geospatial holograms. These results, however, should be further verified in
consecutive studies.
The results from our length code matching algorithm, which will be implemented next,
will provide additional findings on the context of successful geovisualization use. Our future
research will extend this framework and investigate additional spatial thinking processes and
different geovisualization technologies. In addition, it will investigate age related spatial thinking
patterns. Steinke‘s (1987) last sentence on eye-movement analysis research in cartography is
very euphoric: ―let us move forward‖. We could not better summarize it and fully endorse this
statement.
27
Acknowledgements
We would like to thank Roberto Campos, Tobii Technology AB and the thirty-four
undergraduate and graduate students for their voluntary participation in the study. Zebra Imaging
provided the geospatial holograms. Any opinions, findings, and conclusions or recommendations
expressed in this material are those of the authors and do not necessarily reflect the views of the
funding and participating agencies or companies.
28
Literature
Alajlan N. et al. 2006. Shape Retrieval Using Triangle-Area Representation and Dynamic Space
Warping, Technical Report, University of Waterloo.
Andreassi, J. L. 2006. Psychophysiology: Human Behavior and Physiological Response:
Lawrence Erlbaum Associates.
Antes, J. R., K.-T. Chang, and C. Mullis. 1985. The visual effect of map design: An eye-
movement analysis. The American Cartographer 12 (2):143-155.
Arthur, P., and R. Passini. 1992. Wayfinding - People, signs, and architecture. New York:
McGraw-Hill.
Baggs R. A. 1997. Object Matching Using the Dynamic Programming Algorithm, PhD,
Dissertation, Florida Institute of Technology.
Baggs, R. A., and D. E. Tamir. 2008. "Image Registration Using Dynamic Space Warping",
International Conference on Artificial Intelligence and Pattern Recognition, AIPR-08, Florida.
Beatty, J. 1982. Task-evoked pupillary responses, processing load, and the structure of
processing resources. Psychological Bulletin, 91(2), 276-292.
Benton, S. A., and V. M. Bove. 2008. Holographic Imaging. New York: Wiley-Interscience.
Billinghurst, M., and S. Weghorst. 1995. The use of sketch maps to measure cognitive maps of
virtual environments. Paper read at Proceedings of the Virtual Reality Annual Information
Symposium, at Research Triangle Park, NC.
Biocca, F. 1997. The Cyborg's dilemma: Progressive embodiment in virtual environments.
Journal of Computer-Mediated Communication 3 (2).
29
Blades, M. 1991. Wayfinding theory and research: The need for a new approach. In Codnitive
and linguistic aspects of geographic space, edited by D. M. A. U. F. Mark. Dordrecht: Kluwer
Academic Publishers, The Netherlands.
Bollmann, J., M. Johann, and F. Heidmann. 1999. Kartographische Bildschirmkommunikation.
Edited by J. Bollmann. Vol. 13, Beiträge zur kartographischen Informationsverarbeitung. Trier:
Selbstverlag der Geographischen Gesellschaft Trier.
Brown L. G. 1992. A survey of Image Registration Techniques, ACM Computing Surveys, ACM
Press.
Castner, H. W., and J. R. Eastman. 1984. Eye-movement parameters and perceived map
complexity - I. The American Cartographer 11 (2):107-117.
———. 1985. Eye-movement parameters and perceived map complexity. The American
Cartographer 12 (1):29-40.
Committee on the Support for Thinking Spatially. 2006. Learning to think spatially. Washington,
DC: The National Academies Press.
Cornell, E. H., and C. D. Heth. 2000. Route learning and wayfinding. In Cognitive Mapping -
Past, present and future, edited by R. Kitchin and S. M. Freundschuh. London: Routledge.
Dobson, M. W. 1977. Eye Movement Parameters and Map Reading. The American
Cartographer 4 (1):39 - 58.
———. 1979. The influence of map information on fixation localization. The American
Cartographer 6 (1):51-65.
30
Dodge, R. 1900. Visual perceptions during eye movement. Psychological Review 7:454-465.
Downs, R. M., and D. Stea. 1977. Maps in Minds - Reflecting on cognitive mapping. New York:
Harper & Row.
———, eds. 1973. Image and environment. Chicago: Aldine.
Duchowski, A. T. 2007. Eye Tracking Methodology: Theory & Practice. 2 ed. New York:
Springer.
Fabrikant, S. I., Rebich-Hespanha, N. Andrienko, G. Andrienko, and D. R. Montello. 2008.
Novel method to measure inference affordance in static small-multiple map displays representing
dynamic processes. The Cartographic Journal 45 (3):201-215.
Field, M., K. Mogg, and B. P. Bradley. 2004. Eye movements to smoking-related cues: effects of
nicotine deprivation. Psychopharmacology 173 (1):116-123.
Freeman H. 1974. Computer processing of line drawing images, Computer Surveys.
Freundschuh, S. M. 2000. Micro- and macro-scale environments. In Cognitive Mapping - Past,
present and future, edited by R. Kitchin and S. M. Freundschuh. London: Routledge.
Gärling, T., A. Böök, and E. Lindberg. 1986. Spatial Orientation and Wayfinding in the designed
environment - A conceptual analysis and some suggestions for postoccupancy evaluation. J Arch
Plan Res 1986 (3):55-64.
31
Goldberg, J. H., and X. P. Kotval. 1999. Computer interface evaluation using eye movements:
methods and constructs. International Journal of Industrial Ergonomics, 24(6), 631-645.
Grady, R. 1992. Practical Software Metrics for Project Management and Process Improvement:
Prentice-Hall.
Haralick R. M. and G. Shapiro. 2002. Computer and Robot Vision, Vol. I and II, Addison
Wesley.
Hariharan, P. 2002. Basics of Holography. Cambridge: Cambridge University Press.
Heckman, P. 1986. The magic of holography. New York: Atheneum.
Heidmann, F., and M. Johann. 1997. Modelling graphic presentation forms to support cognitive
operations in screen maps. Paper read at International Cartographic Conference, at Stockholm,
Schweden.
Hirtle, S. 1997. The cognitive atlas: Using GIS as a metaphor for memory. In Spatial and
temporal reasoning in geographic information systems, edited by M. J. Egenhofer and R. G.
Golledge. Oxford: Oxford University Press.
Johann, M., and F. Heidmann. 1996. ComTest - Kartographische testsoftware zur empirischen
Untersuchung von Arbeitsprozessen am Bildschirm. Trier: Universität Trier - FB VI/Abt.
Kartographie.
Just, M., A., and P. A. Carpenter. 1976. Eye fixations and cognitive processes. Cognitive
Psychology, 8, 441-480.
32
Kasper, J. E., and S. A. Feller. 2001. The complete hook of holograms - How they work and how
to make them. Mineola, NY: Dover Publications.
Keogh, E. et al. 2006. LB_Keogh Supports Exact Indexing of Shapes under Rotation Invariance
with Arbitrary Representations and Distance Measures, 32nd International Conference on Very
Large Data Bases. Seoul, Korea.
Kitchin, R., and S. M. Freundschuh, eds. 2000. Cognitive Mapping - Past, present and future.
London: Routledge.
Komogortsev, O. V., and U. K. S. Jayarathna. 2008. 2D Oculomotor Plant Mathematical Model
for eye movement simulation. Paper presented at the BioInformatics and BioEngineering, 2008.
Komogortsev, O. V., and J. I. Khan. 2009. Predictive Compression for Real Time Multimedia
Communication Using Eye Movement Analysis. ACM Transactions on Multimedia Computing,
Communications and Applications Journal 5 (1).
Kuipers, B. J. 1978. Modeling spatial knowledge. Cognitive Science 2 (2):129-154.
———. 1983. The cognitive map: Could it have been any other way? In Spatial orientation -
Theory, research, and application, edited by H. L. Pick and L. P. Acredolo. New York: Plenum
Press.
Leigh, R. J., and D. S. Zee. 2006. The Neurology of Eye Movements: Oxford University Press.
MacEachren, A. M. 1991. The role of maps in spatial knowledge acquisition. The Cartographic
Journal 28 (Dezember 1991):151 - 162.
33
———. 1992. Learning Spatial Information from Maps: Can Orientation-Specificity Be
Overcome? Professional Geographer 44 (4):431 - 443.
———. 1995. How Maps Work - Representation, Visualization, and Design. New York: The
Guilford Press.
MacEachren, A. M., and M. J. Kraak. 2001. Research challenges in geovisualization.
Cartography and Geographic Information Science 28(1): 1-11.
McCall, J. A., P. K. Richards, and G. F. Walters. 1977. Factors in Software Quality. : Nat'l Tech.
Information Service.
Nielsen, J. 1993. Usability Engineering. San Francisco, CA, USA: Academic Press.
Peponis, J., C. Zimring, and Y. K. Choi. 1990. Finding the building in wayfinding. Environment
and Behavior 22 (5):555-590.
Péruch, P., L. Belingard, and C. Thinus-Blanc. 2000. Transfer of spatial knowledge from virtual
to real environments. In Spatial Cognition II - Integrating abstract theories, empirical studies,
formal methods, and practical applications, edited by C. Freksa, C. Habel, W. Brauer and K. F.
Wender. Berlin: Springer.
Poole, A., and L. J. Ball. 2004. Eye Tracking in Human-Computer Interaction and Usability
Research: Current Status and Future Prospects. In Encyclopedia of Human Computer
Interaction: Idea Group.
Rayner, K. 1998. Eye Movements in Reading and Information Processing: 20 Years of Research.
Psychological Bulletin 124 (3):372-422.
34
Ruddle, R. A. 2000. Navigating overlapping virtual worlds: Arriving in one place and finding
that you are somewhere else. In Spatial Cognition II - Integrating abstract theories, empirical
studies, formal methods, and practical applications, edited by C. Freksa, C. Habel, W. Brauer
and K. F. Wender. Berlin: Springer.
Salvucci, D. D., and J. H. Goldberg. 2000. Identifying fixations and saccades in eye tracking
protocols. Paper read at Eye Tracking Research and Applications Symposium, at New York.
Shepard, R. N., and S. Hurwitz. 1984. Upward direction, mental rotation, and discrimination of
left and right turns in maps. Cognition 18:161-193.
Siegel, A. W., and S. H. White. 1975. The development of spatial representations of large-scale
environments. In Advances in child development and behaviour, edited by H. W. Reese. New
York: Academic Press.
Software Engineering-Product Quality-Part 1: Quality Model. 2001. Geneva Switzerland:
International Standards Organization.
Sorrows, M. E., and S. Hirtle. 1999. The nature of landmarks for real and electronic spaces. In
Spatial information theory - Cognitive and computational foundations of geographic information
science, edited by C. Freksa and D. M. Mark. Berlin: Springer.
Stea, D. 1976. Program notes on a spatial fugue. In Environmental Knowing - Theories,
Research, and Methods, edited by G. T. Moore and R. G. Golledge. Stroudsburg, PA: Dowden,
Hutchinson & Ross, Inc.
Steinke, T. R. 1987. Eyemovement studies in cartography and related fields. Cartographica 24
(2):40 -73.
35
Swienty, O., T. Reichenbacher, S. Reppermund, and J. Zihl. 2008. The role of relevance and
cognition in attention-guiding geovisualization. The Cartographic Journal 45 (3):227-238.
Tamir, D., O. Komogortsev, and C. Mueller. 2008. An Effort and Time Based Measure of
Usability. Paper presented at the 6th Workshop on Software Quality, 30th International
Conference on Software Engineering.
Thorndyke, P. W., and S. E. Goldin. 1980. Spatial learning and reasoning skill. In Spatial
Orientation - Theory, Research, and Application, edited by H. L. Pick and L. P. Acredolo. New
York: Plenum Press.
Thorndyke, P. W., and B. Hayes-Roth. 1980. Differences in spatial knowledge acquired from
maps and navigation. 1 ed. Santa Monica: The Rand Corporation.
Tlauka, M., and P. N. Wilson. 1996. Orientation-free representations from navigation through a
computer-simulated environment. Environment and Behavior 28 (5):647-664.
Tolman, E. C. 1948. Cognitive maps in rats and men. Psychological Review 55:189-208.
Tversky, B. 1993. Cognitive maps, coginitve collages, and spatial mental models. Paper read at
COSIT 1993, at Marciana Marina, Elba Island.
———. 2000. Levels and structure of spatial knowledge. In Cognitive Mapping - Past, present
and future, edited by R. Kitchin and S. M. Freundschuh. London: Routledge.
Vora, J., S. Nair, A. K. Gramopadhye, A. T. Duchowski, B. J. Melloy, and B. Kanki. 2002.
Using virtual reality technology for aircraft visual inspection training: presence and comparison
studies. Applied Ergonomics 33:559-570.
36
List of Figures
Figure 1: Hologram of the university campus
Figure 2: 1mm x 1mm pixel in a hologram
Figure 3: Information processing model of spatial orientation and way-finding (Gärling et al.
1986)
Figure 4: ―Flat‖ geospatial hologram of San Francisco
Figure 5: Three-dimensional geospatial hologram of San Francisco
Figure 6: Sample fixation points and saccades on a geospatial hologram
37
Figure 1
38
Figure 2
39
Figure 3
Start of travel plan
Termination of travel plan
Formation of travel plan
Execution of travel plan
Decision about destination
Localization of destination
Selection of route
Acquisition of information
about properties of the
environment
Retrieval from
media information
about properties of
the environment
Direct observation
of properties of
the environment
Retrieval/inference
from the cognitive
map information
about properties
of the environment
Maintenance/
restoration of
orientation in the
environment
40
Figure 4
41
Figure 5
42
Figure 6
top related